Here are some interesting articles dealing with using optimisation algorithms to control robots.
The first one is a work done by the LAAS (in Toulouse, yes). The algorithms optimize a trajectory over 5s (at 1ms sampling period) over the whole body of HRP-2.
The article is here and here is a video
It is very interesting since it can predict the complete movement using its whole body and some constraints.
I do the same but not over 5s, only for the next step (to be reactive).
The second one deals with optimizing robot behavior so that it is close to an animtion character. The research is done at Disney Research on 3D printed robots.
The article is here.
And a video:
I remark the ZMP criterion is the only one used to constrain the motion.